what is non-experimental research
When we do not manipulate the independent variable, rather we measure it as it naturally occurs
we can not make causal conclusions in non-experimental research
When is non-experimental reasearch considered more desirable then experimental research
- we are investigating one variable (ie will people obey instructions) - we are not seeing what conditions they will follow instructions under just yes/no so there isn’t the need for two groups just seeing how people respond to one condition
- we are only trying to see if a correlation exists rather then a causal relationship
- we are trying to find a causal relationship but we can not ethically or practically test it (ie if damage to the hippocampus causes issues with memory tracing)
- we are trying to describe a specific scenerio
What are the types of non-experimental studies
correlational research
we are just trying to see if when one variable changes in one direction if the other variable reliably changes in the same or a different direction
Observational research
we are interested in how something occurs in a natural setting without manipulations
ie see how self esteem impacts gpa
have students fill out a questionniare about self esteem and then see their end of year GPAs
is non experimental if you are not manipulating the independent variable or having group random assignment
cross sectional research
we compare people of different age groups at a specific time on a domain
we find people who are 20, 30 and 40 in 2026 and compare them on a domain
con differences may represent the time the people were certain ages in/the world events that occured at thoose times (cohort effects)
longitudinal
we find a group of people who are a certain age and then test them over a domain over a progression of a certain number of years to see how their performance changes with age - we will not have cohort effects here but it is very time consuming
cross sequential
sequ like sequine - this approach shines above the rest - is the best
we make groups of people who are different ages at a specific time and can test them on a domain (this means we get data sooner) and then test how their performance changes over a certain number of years (this allows us to get data without as much influence from cohort effects)
Rank the internal validity of different types of research
Experimental - we control conditons so we can be the most certaint hat our results prove/disprove our hypothesis (rather then something else)
quasi experimental - involves some aspects of experimental research but fails to control some significant extraneous variables ie random assignment
if you want to compare the effectiveness of two different anti-bullying campaigns you can control how the campiagns are administered and how the results are measured however can not randomly assign students to each school. If one school appears to have a greater decrease in bullying it could reflect that that school had the more effective campagin however it could also reflect that the students of that school were just more likely to listen to any campaign (and might have done better if given the other campaign as well) - this makes it so we have less certainity that we can reject or accept our hypthoesis that we saw the change in our dependent variable occur due to the change in the independent variable
non experimental - not doing any manipulation so last sure of what factors cause what to occur
When would we use correlational research
We do not think that one variable causes a change in the other but we do think they are related
We are only interested in describing the relationship (if variable A increases variable B reliably… (different then explain which would specificity if variable A causes the change in variable B or they are both caused by another variable)) or predicting if I do this to variable A this will happen to variable B
When we can not practically or ethically carry out causal research ie alcohol consumptions affect on memory - it is unethical to manipulate participants alcohol consumption -so we have to take data regarding peoples existing alcohol consumption and their data on memory
When we want higher external validity (as external validity increases internal validity decreases)
we can pair up with experimental data so we can be confident of the relationship that exists in real life and have an idea of what might cause it to exist
When we are trying to see how related two things are as a measure of validity and reliabiltiy (reliabile can be done multiple times and get the same results - measures how well the test can get the same results (ie does it have ambiguous questions that may be interpreted differently at different times- low reliablity), whereas validity (if something is valid it can be used-validity measures how well our test can be used to represent our hypothesis)
ie how correlated a participants scores on two different depression tests are. We do not believe taking one of the test creates the score of the other and likely when implemented we are only interested in using one - we just want to make sure they are both reliable.
When is it considered correlational
If two groups exist as two different conditions but the researchers did not manipulate which group participants were in (do not have random assignment to a condition) but rather used the people that already exist in those conditions it is correlational.
what is a positive and negative relationship
positive - it is easier to feel positive when you are surrounded by people who are also uplifting when your uplifted and are not happy when your down (this might feel like they take pleasure in your misery) positive refers to when both varaibles have the change occur in the same direction
negative - opposing views can bring out the naegative side of people - negative relationship - opposing when one increases the other decreases change in different directions
what is pearsons coefficient
measures the strength of a relationship between two variables +1.0 (strongest positive correlation) -1.0 strongest negative correlation. 0 no correlation
+ or - 0.10 considered to represent a weak stregnth correlation, + or - 0.30 considered to represent a medium strength correlation and + or - 0.50 is thought to represent a strong correlation.
a strong corelation will represent a diagnol line either going up from left to right if + bc as x increases so does y and going down from left to right if -
pearsons coefficient only works for linear relationships 0 does not work when a medium ammount is assoicated with one outcome and the two other extremes are associated with another (this creates a U shape graph)
ie the relationship between depression and sleep - find that people who get 8 hours of sleep a night scored the lowest on depression and then the people on the upper and lower extremes scored the highest (bc symptoms of depression can include hypersomnia or insomnia)
what is restriction of range
when we fail to capture the full scope of a certain variable so it skews our perception of the relationship
What does it mean that correlation does not imply causation
occurs for 2 reasons
the directionality problem, we can say that X and Y have a positive relationship (both move in the same direction) however we don’t know if X increasing causes Y to increase or Y increasing causes X to increase
The third variable problem
X and Y might not have a causal relationship with each other - the changes in both of them may be caused by a third variable Z. Spur of the moment decision (a decision we make seemingly on a whim in response/related to an event (so there is some relationship/some correlation) - however there is probably another variable at play that causes this decision) - spuratious correlations are correlations caused by third variables
What is a correlation matrix
matrix like system/web
correlation matrix (system of correlations)
shows each individual variables correlation coefficient with each other variable. ie if we had
need for cognition, intelligence, dogmatism and social desirability as our variables we would have a correlation coefficient for need for cognition and intelligence, need for cognition and dogmatism, and need for cognition and social desirablity (we would do the same for each other variable - so it is just a system where we display multiple different correlations between variables at once- allows us to compare). A variable and itself will always have a correlation coefficient of +1, (as when intelligence increases itself (intelligence) will always increase)
What is factor analysis
when we split aour data into groups where we have variables that are correlated with each other in the same group and variables that are not correalted with each other in different groups. We we then find what underlying domain we think expresses the reason for this correlation and call it the factor.
ie scores on arithmetic, quantitive reasoning and spatial reasoning are positively correlated with each other - so they are clustered together and we explain the factor that connects them as being mathematical intelligence.
Note although the variables in a cluster are positively correalted with each other and not other clusters that doesn’t mean that a person cant fit into two of factors that are said to be represented by different clusters.
Ie liking jazz might be positively correlated with liking blues and soul (making these fit into a cluster which researchers label as upbeat) and not very correlated with liking rock which is correalted and therefore clustered with liking alternative and heavy metal (which has its cluster labelled as intense) however this does not mean that an individual can not like both upbeat and intense music
additionally factor analysis objectively shows what varaibles are correlated together however the throughline that runs through this correlation is judged by the reasercher (the researcher labels a factor for a specific cluster)
What is a partial correlation
Used when we want to determine the relationship between two variables but we suspect that a third variable may be responsible for creating at least part of the relationship partial in that we want to get the part of the total relationship of variable A and B that is not influenced by variable C. We adjust for variable C using statistical control to get the relationship between variable A and B independent of variable C. We can not draw causal conclusions because we still have the directionality problem (we don’t know if variable A is impacting variable B or if variable B is impacting variable A) also since we are not testing in a setting where we have control over lots of variables there still could be other third party variables creating the relationship between variable A and B
what is a single regression
in order to do a regression we must have the correlation mapped onto a graph (so we must know the correlation). Then we can predict the value of a given varaiable based on the value of another variable.
one regression formula: Y = b1X1
b1 = the slope, (aka the regression weight) X1 = the value of the variable that we know (our predictor variable), Y = the value of the varaible that we are trying to predict/our predicted value, (Y = our criterion or outcome variable)
what is a multiple regression
when we have multiple predictor varaibles that we are trying to determine an outcome variable from.
formula : y = b1X1 + b2X2 + b3X3
where b = the amount that- the X variable contributes on average to the outcome variable
pros this allows us to see which predictor variable has the most influence on the outcome variable and allows us to see the influence of each predictor variable independent from each other.
What are the types of interviews used in qualitative analysis
Focus groups
involves having a small group of people interviewed at once
cons
- people may change their answer to fit with others bc they are afraid of being judged
- some peoples opinons may be more represented bc they might be more extraverted and dominate the discussion
unstructured interviews
no script is followed
can be good for sensitive topics when the participants should get to dictate how much they want to share
structured interviews
researcher follows an exact script to what questions they will ask
semi structured
researchers follow a script to an extent and then abandon the script and ask questions based on participants answers to the earlier bits of the script
What is grounded theory
our ground that we start qualitative research from is the last point in our research collection in quantitative research
quantitative - theory - hypothesis - collect data
qualitative - examine data - discern what themes show up in the data and form a theory, (an interpretation of what these themes mean (called the theoretical narrative))
the theoretical narrative is subjective based on the researchers opinon not a universal measure)
Qualitative vs Quantitative research
Qualitative
- provides more specific information about a smaller group of people
- does not involve numerical measures
- can use grounded theory where we start with the data, researchers then identify what ideas are repeated in the data and interpret what themes these repetitions reflect - the researchers then form a theory based on their these themes (called a theoretical narrative)
- since we start with data instead of answering a specific question we can form interpretations that give us opportunities to generate new questions
pros: rich detail about what a specific experience is like
cons: low objectivity, validity and reliability
Quantitative
- uses numerical measures
- start with a theory, form a hypothesis and then gather data
- provides more vague information that is applicable to a larger group of people (uses larger sample sizes)
pros: high objectivity, validity and reliability
cons: may not capture all of the richness of an experience
What is mixed methods and what are two approaches that can be used under it
mixed methods approach uses both qualitative and quantitative methods
can use qualitative research to generate a hypothesis and quantitative research to test the hypothesis
triangulation- like an equilateral triangle have the two lines that meet at a point represent qualitative and quantitative research and the line that connects them be the question. Involves doing both qualitative and quantitative research simultaneously and seeing if they arrive at the same point (conclusion)
What is a factorial design
involves seeing how more then one independent variable impact a dependent variable
In order to see the total different number of conditions you need to fufill for a factorial design you multiply the number of different possible expressions for each varable together. (so the numbers listed represent the number of ways their varaible can be expressed and the ammount of numbers we have being multiplied = the ammount of independent variables we are examining)
ie 3 x 2 x 2
3 means the first independent variable can be expressed in 3 ways (ie morning, midday, night)
2 means the second independent variable can be expressed in 2 different ways (ie no music, music)
2 means the third independent variable can be expressed in 2 ways (ie on the phone, not on the phone)
since we have 3 numbers being multiplied together tells us we have 3 independent variables
can do between subjects factorial designs (where the condition that is being represented differs between subjects- each person represents one condition only) or within subjects factorial designs (where each condition is represented within the subject - each subject represents all conditions) in a factorial design
we can also choose to have one independent variable be done following within subjects design (where we subject, subjects to all expressions of that variable) and another one done as between subjects design (where we only subject each subject to one - not all expressions of that variable) within the same trail since we have multiple variables in factorial designs (this is called a mixed factorial. design)
What are the 3 things that occur when we have a factorial design where one of our independent variables is not manipulated
- Our non manipulated independent variable usually has to be a between subject variable (is a binary category that the participant already exists in before the experiment - ie hypochondriac or not). Since our non manipulated independent variable is binary and its expression is not created by us it is a between subject variable - ie can’t make someone who was a hypochondriac beforehand suddenly not be a hypochondriac for our other condition
- So long as we have one independent variable that is manipulated it is still considered an experiment
- We can only draw causal conclusions about the independent variable that we manipulate (the pre-existing independent variable that we do not manipulate (and likely can not manipulate)’s relationship to the dependent variable can not be judged to be causal as their could be a third factor that creates both the non-manipulated independent variables expression and the dependent variables expression
ie testing to see peoples memory of a list of words
manipulated independent variable: if the words mentioned are negative medical words (ie tumor) or not
non-manipulated indepdent variable: if the person is a hypochondriac or not
dependent variable: how much of the list the person remembers
we would have 2 x 2 conditions
if the people who are given the negative medical words remember more of the words we can say that being given negative medical words causes individuals to remember the words better - however we can not say if being hypochondriac or not influences the likelihood of them remembering the list better as it could be that hypochondria does not cause this better memory and rather that a greater level of mental review causes both hypochondria and better memory
can we have a non - manipulated factorial design
yes we can have a factorial design where none of our independent variables are manipulated however we can not draw causal conclusions
How do we represent factorial designs on graphs
the y axis is always for the dependent variable (the big y like question (the one who will be the answere to our question (how the dependent variable is manipulated))
we can represent the two variables having the x axis represent one of them and then having two seperate points or bars for that position whose colors represent the other independent variable
What are main effects
when we take the average impact of one variable expression across all conditions.
ie we have
no phone driving during the day
no phone driving at night
phone driving during the day
phone driving at night
we want to see the difference between the average of our performance if we use the phone or not (we want to see the main effect of the difference of using a phone or not- the average difference across all of our performances)
